Multi-objective Optimization of Dependability Attributes Using An Asynchronous Heterogeneous Hierarchical Parallel Genetic Algorithm

The optimal satisfaction of dependability attributes and cost, in the design of engineering systems, is a hard multi-objective optimization problem which requires automated algorithms that can effectively search large design spaces. In this paper, a recently proposed Asynchronous Heterogeneous Hierarchical Parallel Genetic Algorithm (AHHPGA) is extended with multi-objective optimization capabilities and applied to a particular, benchmark, redundancy allocation formulation of the problem that takes into account the reliability, cost and weight of a system. A new model-based safety and reliability evaluation technique is also employed to overcome the limitations imposed by the use of Reliability Block Diagrams in earlier work.